|
on Forecasting |
| By: | Sylvia Kaufmann (Study Center Gerzensee) |
| Abstract: | We document whether a simple, univariate model for quarterly GDP growth is able to deliver forecasts in a crisis period like the Covid-19 pandemic, which may serve cross-checking forecasts obtained from elaborate and expert models used by forecasting institutions. We include shocks to the log number of short-time workers as timely available current-quarter indicator. Yearly GDP growth forecasts implied by quarterly forecasts serve cross-checking, in particular at the outbreak of the pandemic. |
| Date: | 2022–02 |
| URL: | http://d.repec.org/n?u=RePEc:szg:worpap:2202&r= |
| By: | Massimo Guidolin; Davide La Cara; Massimiliano Marcellino |
| Abstract: | With reference to S&P 500 daily returns, we report evidence of an in-sample predictive accuracy breakdown for realized variance by GARCH models in correspondence to the March 2020 Covid-19 outbreak. However, a variety of macroeconomic risk, political and social media sentiment uncertainty factors, and crucially a few variables capturing the evolution of the Covid-19 pandemics, successfully predict the direction and size of GARCH forecast errors between November 2019 and June 2020. Predictors related to diagnosed cases, their rate of growth, and the progression of the curve of deceased, infected people in the United States are featured prominently. We test a number of “augmented” GARCH models to include the most precisely estimated exogenous variables and find that they offer precise forecasts in samples that include the Covid-19 outbreak. In genuine out-of-sample tests, augmenting GARCH with Covid-19 related exogenous variables increases the percentage of days in which the direction of change in realized variance is correctly predicted. |
| Keywords: | Conditionally heteroskedastic models, Covid-19, volatility forecasting |
| JEL: | C32 C53 E47 G01 |
| Date: | 2021 |
| URL: | http://d.repec.org/n?u=RePEc:baf:cbafwp:cbafwp21169&r= |
| By: | Iuri H. Ferreira; Marcelo C. Medeiros |
| Abstract: | In this paper we examine the relation between market returns and volatility measures through machine learning methods in a high-frequency environment. We implement a minute-by-minute rolling window intraday estimation method using two nonlinear models: Long-Short-Term Memory (LSTM) neural networks and Random Forests (RF). Our estimations show that the CBOE Volatility Index (VIX) is the strongest candidate predictor for intraday market returns in our analysis, specially when implemented through the LSTM model. This model also improves significantly the performance of the lagged market return as predictive variable. Finally, intraday RF estimation outputs indicate that there is no performance improvement with this method, and it may even worsen the results in some cases. |
| Date: | 2021–12 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2112.15108&r= |
| By: | Fabio Ashtar Telarico (CSEES, FDV) |
| Abstract: | Tax analysis and forecasting of revenues are of paramount importance to ensure fiscal policy's viability and sustainability. However, the measures taken to contain the spread of the recent pandemic pose an unprecedented challenge to established models and approaches. This paper proposes a model to forecast tax revenues in Bulgaria for the fiscal years 2020-2022 built in accordance with the International Monetary Fund's recommendations on a dataset covering the period between 1995 and 2019. The study further discusses the actual trustworthiness of official Bulgarian forecasts, contrasting those figures with the model previously estimated. This study's quantitative results both confirm the pandemic's assumed negative impact on tax revenues and prove that econometrics can be tweaked to produce consistent revenue forecasts even in the relatively-unexplored case of Bulgaria offering new insights to policymakers and advocates. |
| Date: | 2021–12 |
| URL: | http://d.repec.org/n?u=RePEc:arx:papers:2112.15431&r= |
| By: | Nurdaulet Abilov (NAC Analytica, Nazarbayev University); Aizhan Bolatbayeva (NAC Analytica, Nazarbayev University) |
| Abstract: | In this paper, we use the modified expectation-maximization algorithm of Banbura and Modugno (2014) to estimate a factor model using an incomplete and mixed-frequency dataset for Russia. We estimate and check the forecast accuracy of factor models that differ in the number of factors, the lag structure of the factors, and the presence of autocorrelation in the idiosyncratic component. We choose the best model using the root mean squared forecast error and use the model to compute news contributions to forecast revisions of GDP growth in Russia around crisis periods. We find that the benchmark model with a medium-size dataset and four factors outperforms all other versions of the factor model, simple AR(1) and random walk models. The news contributions to GDP growth revisions around economic downturns in Russia show that the benchmark factor model is extremely good at capturing the impact of new data releases on GDP growth revisions. |
| Keywords: | Factor model; EM-algorithm; Nowcasting; Business cycle index; Russia. |
| JEL: | C53 C55 E32 E37 |
| Date: | 2021–12 |
| URL: | http://d.repec.org/n?u=RePEc:ajx:wpaper:18&r= |